ML flags kidney decline in T2D
New machine‑learning models were reported to predict kidney function decline in Type‑2 diabetes patients — a tool that could trigger earlier nephrology referral or renoprotective therapy. The update came through recent social reporting highlighting model performance and potential clinical rollout. (x.com)
A Scientific Reports preprint published March 27, 2026 describes machine‑learning models built to predict future estimated glomerular filtration rate (eGFR) in people with type 2 diabetes. (nature.com) The authors trained and tested models on 974 patients with a median follow‑up of 5.3 years, using 54 baseline clinical features collected during routine medical checkups and a separate annual time‑point variable. (nature.com) Three algorithms were compared—Light Gradient Boosting Machine, Random Forest, and support vector machine (SVM)—and the SVM produced the best overall fit with an R2 of 0.67 (95% CI 0.62–0.72) across a wide eGFR range, maintaining moderate or better performance for up to six years. (nature.com) The paper reports that a standard multiple linear regression (MLR) benchmark lost predictive accuracy beyond four years, and the authors note the models were developed using only baseline, routinely available data—an approach the team says could facilitate primary‑care integration. (nature.com) The manuscript states the underlying datasets are available from the corresponding author on reasonable request, indicating no public data dump at publication. (nature.com) For comparison, a previously published externally validated prognostic model that pooled three multinational cohorts (n=4,637) and was summarized in 2023 reported R2 values from about 0.70 at year 1 to 0.58 at year 5, a 5‑year C statistic near 0.79, and was released as a public web application. (ajmc.com)
